COVID-19 Crisis Market Updates
The methodology used in the COVID-19 analysis is based upon methods discussed in Valuation and Risk Management in Energy Markets, Glen Swindle (Cambridge University Press, 2014). Weather variables, drift and seasonal components represented as Fourier terms are estimated, with variables retained selected by an out-of-sample estimation criterion. Regressions are performed at the delivery bucket level (e.g. 5×16 for peak). The resulting residuals are then corrected for biases resulting from short time-scale effects to account for notable holiday periods.
In this analysis the data used for model calibration spanned 2016 to mid-November 2019. All residuals (departures from normal) after 18Nov2019 are out-of-sample estimates constructed by subtracting loads as estimated using realized weather and seasonal terms from the realized loads.